Wine 11.10 is out with vkd3d 2.0, VBScript compatibility improvements and more

The latest development release of the Windows compatibility layer Wine 11.10 has arrived with new features and bug fixes. We usually see one new main stable release of Wine per year, so expect Wine 12 sometime in January 2027. Valve currently have Proton 11 that's based on Wine 11 in Beta for Steam gaming.

Here's the main highlights for Wine 11.10 (changelog):

• Bundled vkd3d upgraded to version 2.0.
• XPath support reimplemented without libxml2.
• More VBScript compatibility improvements.
• Various bug fixes.

17 bugs were noted to be solved including issues with Star Wars Racer, Age of Empires III, Star Wars: Knights of the Old Republic, Photolemur 3, Vocaloid 6 and more.

For vkd3d 2.0, which is Wine's own Direct3D to Vulkan translation library that was released earlier in May, there's a lot of big changes there too for new feature support:

libvkd3d

• Resources with a combined depth/stencil format like DXGI_FORMAT_R32G8X24_TYPELESS can be created without the D3D12_RESOURCE_FLAG_ALLOW_DEPTH_STENCIL flag being set. It's somewhat rare for applications to create such resources; typically resources with a combined depth/stencil format are either used with both a depth/stencil view and a shader resource view, or exclusively with a depth/stencil view.

• The syntax for the VKD3D_DEBUG environment variable has been extended to allow more precise control over the debug output. See the ‘Environment variables’ section of the README for a description of the extended syntax.

libvkd3d-shader

• New features and improvements for the HLSL source type:

  • Store operations to structured thread group shared memory.
  • Interlocked operations on structured and/or arrayed unordered access views and thread group shared memory.
  • A new common subexpression elimination optimisation pass.
  • Initial support for loops in shader model 2-3 target profiles.
  • More efficient usage of temporary registers, primarily due to improvements to the register allocator and the introduction of an output write hoisting pass. This is particularly relevant for shader model 1-3 target profiles, where the number of temporary registers is relatively limited, and we may otherwise not be able to compile some shaders.
  • The tex3Dbias(), tex3Dlod(), and texCUBElod() intrinsic functions.
  • The ‘SV_ClipDistance’ and ‘SV_CullDistance’ input/output semantics.
  • The ‘SV_StencilRef’ pixel shader output semantic.
  • Corrected handling of InterlockedMin() and InterlockedMax() operations on inputs with differing signedness. When either of the inputs has an ‘uint’ type, the other is converted/promoted to ‘uint’ as well.
  • The VKD3D_SHADER_COMPILE_OPTION_BACKCOMPAT_MAP_SEMANTIC_NAMES flag now additionally maps the shader model 3 ‘VFACE’ and ‘VPOS’ semantic names to their shader model 4+ equivalents.
  • Constant folding of floating-point modulo expressions. I.e., ‘x % y’ where ‘x’ and ‘y’ are floating-point constants.
  • The scope of variables declared in ‘for’ loop initialisation clauses has been expanded to extend beyond the end of the loop. I.e.,
    ‘for (int i = 0; i < 10; ++i); return i;’ is valid HLSL. That's different from e.g. C99, where the scope of ‘i’ would be limited to the loop.
  • Floating-point literal parsing has been made independent of the current locale. Previously, the HLSL parser would use the current locale's decimal separator when parsing floating-point literals, instead of the HLSL decimal separator, which is always the ‘.’ character. Note that this only affects programs that set a locale other than the initial ‘C’ locale, for example by calling the setlocale() function.

• New features and improvements for the effects (FX) source type:

  • Parsing of operand indices in eval() block was broken and has been fixed.
  • The ‘umod’ FXLVM operation is recognised.

• New features and improvements for the legacy Direct3D byte-code source type:

  • The following instructions:
    • m4x4, m3x4, m4x3, m3x3, and m3x2
    • phase
    • texdepth
    • texreg2ar, texreg2gb, and texreg2rgb
  • The ‘_dz’, ‘_db’, ‘_dw’, and ‘_da’ source modifiers. These are used with the ‘texcrd’ and ‘texld’ instructions.
  • The ‘vFace’ and ‘vPos’ pixel shader input registers.
  • Shader model 3 vertex shader point size outputs. In shader model 1 and 2 point size output uses the ‘oPts’ output register; shader model 3 uses a generic ‘o’ register with a ‘dcl_psize’ declaration.
  • Shader model 1 and 2 vertex shader outputs are clamped to the 0 to 1 range.

• The DXIL source type supports forward-referencing pointers in load, store, atomic read/modify/write, and compare/exchange operations.

• The DXIL source type supports pixel shader specified stencil reference values.

• The experimental Metal Shading Language (MSL) target supports pixel shader specified stencil reference values.

• The Direct3D shader assembly target supports 16-bit immediate constants. These are typically produced by DXIL sources.

• The Direct3D shader assembly target supports the following global flags:

  • ‘64UAVs’
  • ‘ROVs’
  • ‘UAVLoadAdditionalFormats’
  • ‘UAVsAtEveryStage’
  • ‘allResourcesBound’
  • ‘enable11_1ShaderExtensions’
  • ‘int64Ops’
  • ‘nativeLowPrecision’
  • ‘stencilRef’
  • ‘viewportAndRTArrayIndex’
  • ‘waveOps’

• New interfaces:

  • The vkd3d_shader_scan_denormal_mode_info structure extends the vkd3d_shader_compile_info structure, and can be used to retrieve the denormal mode used for floating-point numbers.
  • The VKD3D_SHADER_SPIRV_EXTENSION_KHR_FLOAT_CONTROLS enumeration value indicates support for the SPV_KHR_float_controls extension in the SPIR-V target environment.
  • The VKD3D_SHADER_COMPILE_OPTION_DENORMAL_MODE_F16 compile option specifies the denormal mode to use for 16-bit floating-point numbers.
  • The VKD3D_SHADER_COMPILE_OPTION_DENORMAL_MODE_F32 compile option specifies the denormal mode to use for 32-bit floating-point numbers.
  • The VKD3D_SHADER_COMPILE_OPTION_DENORMAL_MODE_F64 compile option specifies the denormal mode to use for 64-bit floating-point numbers.
  • The VKD3D_SHADER_COMPILE_OPTION_CONST_GLOBAL_UNIFORMS flag specifies that all uniforms with global scope should be considered ‘const’ in HLSL sources.
  • When targeting VKD3D_SHADER_API_2_0, compilation will fail when a required floating-point denormal mode can't be specified in the target shader. For the SPIR-V target, this requires the SPV_KHR_float_controls extension, as well as the corresponding capabilities in the target environment. The earlier mentioned compile options allow overriding the denormal modes specified by the source shader.

• The syntax for the VKD3D_SHADER_DEBUG environment variable has been extended to allow more precise control over the debug output. See the ‘Environment variables’ section of the README for a description of the extended syntax.

libvkd3d-utils

• The D3DCompile(), D3DCompile2(), D3DCompile2VKD3D(), and D3DPreprocess() functions support using D3D_COMPILE_STANDARD_FILE_INCLUDE as include handler.

• Unless the D3DCOMPILE_ENABLE_BACKWARDS_COMPATIBILITY flags is specified, D3DCompile2VKD3D() will consider all uniforms with global scope to be ‘const’ for compiler versions 37 and up. This is also the case for D3DCompile() and D3DCompile2(), which behave like compiler version 47.